I recently stumbled across this set of photo composites that have been floating around the web for the past year, featuring a variety of celebrities inserted into someone’s party photos. What impressed me about these photos was that the creator did an unusually good job of making them look — at least at first glance — plausible. I thought these photos would be a good test of some of our photo forensic techniques.
Welcome to the Fourandsix blog, where you’ll find tips on image forensics techniques and commentary on issues relevant to photo tampering and the responsible use of imaging tools.
Go outside and stand with your back towards the sun and look at your shadow that is cast onto the sidewalk. You will notice that the shadow near your feet has a sharp boundary, while the shadow near your head is fuzzy. This simple observation regarding the expect shape of a shadow can be used to detect fake shadows in an image.
What is the standard by which you judge whether a photo has been over-manipulated? Do you judge it against the reality of the original scene that was photographed? Certainly that makes sense, but it’s hard to judge that objectively if you weren’t present when the photograph was captured. Instead, it’s common to judge the modifications against the original, unaltered photograph that the camera captured—despite the fact that the camera itself tends to change the scene in subtle ways due to lens distortion, sensor design, and internal processing algorithms. A new technology trend known as computational photography, however, is changing how we approach image capture, and this may have implications for how we judge what is manipulated.
Grooves made in a gun barrel impart a spin to the projectile for increased accuracy and range. These grooves introduce distinct markings to the bullet fired, and can therefore be used to link a bullet to a specific handgun. In the same spirit, imperfections in manufacturing lead to slight deviations between the precise amount of light that strikes a camera sensor and the recorded pixel values. These deviations vary from camera to camera (even of the same make and model) and can therefore be used for camera ballistics — linking an image to a specific digital camera.
A power grid delivers power at a specific frequency termed the Electric Network Frequency (ENF). The ENF varies from country to country, but each power grid strives to deliver power at a fixed frequency. Due to varying loads on the power grid the ENF fluctuates slightly, typically by less than a fraction of a percent. These fluctuations affect digital recording devices enough so that the fluctuations can be seen in the recording. Recent advances have shown how ENF fluctuations can be used to authenticate digital recordings, particularly audio recordings.